Comparing Microfluidic Performance of Three-Dimensional (3D

Subscriber access provided by University of Newcastle, Australia

Technical Note

Comparing microfluidic performance of 3D printing platforms Niall P. Macdonald, Joan Marc Cabot, Petr Smejkal, Rosanne M Guijt, Brett Paull, and Michael C. Breadmore Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b00136 • Publication Date (Web): 10 Mar 2017 Downloaded from http://pubs.acs.org on March 12, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Comparing microfluidic performance of 3D printing platforms Niall P. Macdonald,a,b Joan M. Cabot,a,b Petr Smejkal,b Rosanne Guijt,c Brett Paull,a,b, and Michael C. Breadmore*,a,b a ARC Centre of Excellence for Electromaterials Science, School of Physical Sciences, University of Tasmania, Sandy Bay, Hobart 7001, Tasmania, Australia. b Australian Centre for Research on Separation Science, School of Physical Sciences, University of Tasmania, Sandy Bay, Hobart 7001, Tasmania, Australia. c Pharmacy School of Medicine, University of Tasmania, Hobart 7001, Tasmania, Australia * Email: [email protected]

ABSTRACT: Three dimensional printing (3D printing) has emerged as a potential revolutionary technology for the fabrication of microfluidic devices. A direct experimental comparison of the three 3D printing technologies dominating microfluidics was conducted using a Y-junction microfluidic device; the design of which was optimized for each printer - fused deposition molding (FDM), Polyjet, and digital light processing stereolithography (DLP-SLA). Printer performance was evaluated in terms of feature size, accuracy, and suitability for mass manufacturing; laminar flow was studied to assess their suitability for microfluidics. FDM was unsuitable for microfabrication with minimum features of 321 ± 5 µm, and rough surfaces of 10.97 µm. However for microfluidic devices > 500 µm, rapid mixing (71 ± 12% after 5 mm, 100 µL/min) indicate strengths in fabricating micromixers. Polyjet fabricated channels with a minimum channel size of 205 ± 13 µm, with surface roughness of 0.99 µm. Despite this, increased mixing (27% ± 10%) suggested suitability for microfluidics where flow splitting would not be a requirement, such as cell culture or microdroplet generators. DLP-SLA fabricated a minimum channel size of 154 ± 10 µm, and 94 ± 7 µm for positive structures such as soft lithography templates, with roughness of 0.35 µm. These results in addition to low mixing (8% ± 1%) showed suitability for microfabrication, and microfluidics requiring precise control of flow. Through further discussion of the capabilities, and limitations of these printers, we aim to provide guidance toward the selection of the 3D printing technology most suitable for specific microfluidic applications.

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Additive manufacturing, or 3D printing, has gained significant attention in manufacturing because it can directly create 3D designs from a computer aided drawing (CAD) file. For more than a decade, engineers and designers have been using 3D printers to make prototypes quickly and cheaply before embarking on the expensive business of equipping a factory to produce the final product. Over time, 3D printing technologies have advanced and have been adapted to work with a broad range of materials, including production-grade plastics and metals. Consequently, the technology evolved from rapid prototyping approach to a manufacturing method with more than 20% of the output in the form of final products1. 3D Printing has found applications in the manufacture of planes, shoes, and clothing, but also in the medical field providing bone replacements and dental implants. There has been a significant effort towards exploring the potential of 3D printing for new areas of applications, such as microfluidics, and biomedical engineering. While 3D printing was first developed in 1980’s2, there has been a surge of interest in the implementation of 3D printing in non-traditional areas, with 80% of the academic papers on 3D printing published in 2012 and beyond. In the field of microfluidics, 3D printing offers the capability to directly print complex 3D microfluidic devices with lowcost desktop printers, changing the way in which such devices

Page 2 of 10

are conceived, designed and manufactured. This underpins a potential rapid growth and expansion of industries involved in the manufacturing of portable chemical sensors, diagnostic and biomedical research devices. Providing design and the evaluation of prototype devices in hours at the cost of tens of dollars per device or less, 3D printing provides an attractive alternative to the weeks and hundreds to thousands of dollars typical for traditional microfabrication approaches. In the late 90’s, the introduction of PDMS and soft lithography allowed researchers to prototype with simple and low-cost infrastructure, significantly expanding microfluidic research. However, PDMS has many known limitations3 including the complicated commercial translation of academic research due to difficulties in large volume manufacturing. To this end, 3D printing adds the potential to prototype in materials that are compatible with large volume manufacturing such as thermoplastics used in embossing and injection molding to the simplicity, flexibility and cost of soft lithography, and may even be a viable manufacturing approach by itself for highly complex 3D geometries, or specialist devices that may only require low to moderate volume production. The suite of 3D printing technologies has been discussed in detail4,5, of particular interest for microfluidics are fused deposition molding (FDM), stereolithography (SLA), inkjet/Polyjet, two photon lithography, selective laser sinter-

Figure 1. Illustration of 3D fabrication methods, and photographs of 3D printed laminar flow devices with 500 µm input channels, and 750 µm laminar flow channel filled with green food dye. A) FDM method in which molten plastic is extruded through a heated nozzle according to the G-code. The features are formed by moving the nozzle in the XY plane until the current layer was complete, when the build platform was then dropped by a set level (50 µm) before moving to the next layer. B) Polyjet (i3DP) method where two sets of 4 micro nozzle arrays, build and support material respectively, spraying micro droplets of polymer to form the device. Following each pass, UV lamps polymerise the material before the layer is levelled by a roller and scraper. C) DLP-SLA method in which 365 nm light was projected onto a build platform immersed in liquid photopolymer. Following each exposure, the build platform was raised by 50 µm then lowered back into the resin bath for the next layer. Scale bars are 1 cm.


Page 3 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


ing, and layered hydrospinning.6 FDM was first demonstrated as a method for fabricating templates for soft lithography of microfluidic devices with PDMS in 2002 7, which was followed shortly thereafter by the same group using an inkjet 3D printer. In 2012, FDM printing was used for the direct fabrication of microfluidic chemical reaction-ware 8. As the resolution and accessibility of 3D printers has evolved, so too has the complexity of 3D printed microfluidic devices, with recent reports demonstrating direct printing of 3D micro-mixers9 and devices with integrated valves 10. We have shown that 3D printed channels of 250 x 250 µm can be achieved by DLPSLA in a matter of minutes using commercially available desktop 3D printers 9 and using DLP-SLA with an in-house developed resin Nordin et al fabricated enclosed microchannels with dimensions less than 100 µm (60 x 108 µm)11, as well as integrated valves, pumps and multiplexors.12 3D printing technology has also found its way into soft-robotics with the autonomous microfluidic powered soft octopus 13, and our group reported laser sintering for the fabrication of 3D metallic microfluidic devices 14. As summarized above, a variety of 3D printing approaches has been used for the fabrication of microfluidic devices, documented in over 100 papers related to 3D printing and microfluidics. This includes a number of reviews that discuss and contrast the different printing approaches and the devices that have been made by each6,15-19. With a wide range of printers on the market, the suitability for the microfluidic market has been studied by comparing different SLA printers20, different inkjet printers both from Stratasys, and 3D Systems 21, as well as a direct comparison between Polyjet and FDM technologies22. While these are all valuable contributions, a cross platform comparison including microfluidic performance is missing. Here, the three most commonly used printing technologies in microfluidics –FDM, Polyjet and DLP-SLA – are critically compared for their ability to fabricate microfluidic devices. The design of a Y-junction microfluidic device was optimized for each printer to achieve optimal microfluidic performance. In comparing the devices, consideration was given to the positive and negative feature analysis of SEM images, measurement of surface roughness, production and infrastructure requirements, and a microfluidic performance based on a study of mixing two adjacent flows within the channels. This report aims to provide the information required for informed decision making, as compromises will be required when selecting a 3D printing approach in microfluidics. In a field as diverse as microfluidics, there is not a single 3D printing technology capable of accommodating all the needs, but a technology that best serves each individual application. Experimental Materials. Veroclear-RGD810 print material and SUP707™ water-soluble support were purchased from Stratasys Ltd. (Minnesota, USA). Crystal Clear acrylonitrile-butadienestyrene (ABS) 1.75 mm diameter filament was purchased from 3D Printing Systems (Melbourne, Australia). BV-007 photopolymer was purchased from MiiCraft (Young Optics Inc, Hsinchu, Taiwan). Isopropanol, and sodium hydroxide (NaOH) were purchased from Sigma Aldrich (Sydney, Australia). Colored food dyes were purchased locally and used as received. Deionised water was provided by a Merck Millipore purification system (MA, USA).

Instrumentation. The laminar flow devices were designed using SolidWorks 2015-2016 (Dassault Systèmes SE, France), and were printed with a Felix 3.0 (FELIX 3D printers, Nieuwegein, The Netherlands), Objet Eden 260VS professional 3D printer (Stratasys Ltd., Minnesota, USA), and a MiiCraft+ desktop DLP-SLA 3D printer (Young Optics Inc, Hsinchu, Taiwan). Slicing of .stl files into G-code for the Felix 3.0 was completed using KISSlicer v1.5, and printing controlled by Repetier Host v1.05 (Willich, Germany). Objet Suite v9.211.3626 (Stratasys Ltd, Minnesota, USA), sliced and processed models for printing as according to manufacture guidelines. For the MiiCraft+, devices were sliced using Creation Workshop (DataTree3D, Dallas, Texas, USA), and edited in Photoshop Elements 14.1 (Adobe Systems Inc, California, USA) before printing on the MiiCraft+ controller software. Surface data was collected using a Veeco Wyko NT9100 Surface Profilometer (Bruker Corporation, Billerica, Massachusetts, USA), data processed with Vision 4.20. SEM images were taken using an Analytical UHR Schottky Emission Scanning Electron Microscope SU-70(Hitachi, Chiyoda, Tokyo, Japan). For pumping of dye into the devices, a dual syringe pump (Harvard Apparatus Inc, MA, USA) with disposable 10 mL plastic was used, with sample volumes of 10 mL for both blue, and yellow dyes. Flow rates selected were 25, 50, and 100 µl/min. PEEK tubing OD 1.59, ID 0.508 mm (P/N: 052308) was supplied by Thermo Fisher Scientific (MA, USA). For measurement of fabricated channels and diffusion cones, ImageJ was used. Fabrication of laminar flow devices. Fabrication of devices used three methods: (i) FDM (Felix 3.0), (ii) PolyJet (Eden 260VS), (iii) DLP-SLA (MiiCraft+) shown in Figure 1. Test parts containing open and closed channels with 1000, 750, 500, 350, and 250 µm dimensions, with 1:1 ratio and 1:2 ratio channels were printed for analysis and characterization. Due to the limitations in the resolution that can be obtained using the Felix 3.0, only devices containing 1000, 750, and 500 µm channels were produced for comparison with the other printers. All structures were printed and analyzed in triplicate. FDM printing. Extruding thermal plastic through a heated nozzle, which was controlled by two precision stepper motors following coordinates specified by G-code generated using KISSlicer, using the following parameters: 100 µm layer height, extrusion width 300 µm, infill 100%, 5 mm/s print speed, 1 loop stroke. After depositing each layer, the build platform was lowered 100 µm before extruding the next layer. A single nozzle was used to extrude Crystal Clear ABS at a temperature of 210 °C for the first layer, 185 °C for the rest while the build platform was at 70 °C throughout the print. After completion of the print run, reservoirs were cleared of stray fibers. No support material was used. PolyJet printing. Jetting a photopolymer using linearly arranged nozzles, spraying micro droplets onto the build surface where the material was polymerized using an integrated UV light source. Voids in the model were filled with a support material, which was removed after completing the printing process. All devices were fabricated with Veroclear-RGD810 build material in combination with SUP707™ as water-soluble support. Orientation of the devices was always in line with the print head so that the roller texture was parallel to the main fluidic microchannel, and printed flat on the build platform in



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

matte mode. To dissolve the soluble support devices were soaked in water for up to 6 h depending on the channel size, followed by soaking in 2% NaOH for up to 3 days with sonication, and finally flushed by a water jet. DLP-SLA printing. Optically curing a photopolymer through a clear bottomed bath by a digital projector, with the vertical movement of the build platform controlled by a single stepper motor. Using the projector, each slice of the device was projected onto the bottom the bath to attach to the build platform. After each exposure, the platform with the previous layer(s) was raised, then lowered to a position minus the height of the previous layer (50 µm in this work). Post-processing of the device involved removal of non-polymerized resin from the channels using compressed air and bonding the device to a glass microscope slide (25 x 75 mm) by pressing a glass slide to the device onto which a drop of BV-007 resin was placed, before curing by exposure to UV light (365 nm) for 5 min. This bonds the device to a glass base for easier handling, and post-cures the structure. Results and Discussion Three printers – FDM, DLP-SLA and Polyjet – were selected for evaluation. The manufacturer specifications are provided in Table 1, and schematic representations of the printing processes are shown in the first row of Figure 1 as well as photographs of fabricated devices from each of the printers.

Table 1. Comparison of fabrication, and production factors of microfluidic devices for the 3D printers in this work

Microfabrication. Initial characterization of the three printers was performed by printing a series of positive and negative structures data show in Table 1, and Figure 2 showing SEM images a small piece of the template that has both 250 and 350 µm open and closed channels. The channels from the FDM

Page 4 of 10

printer (a), did not form according to the design. There were polymer beads making connecting channels (1-2) and rough and deformed areas were the nozzle changed direction (3). Channels printed with the Polyjet printer (b) met the 250 and 350 µm channels specified in the design (4-5). The spray and roll process of the PolyJet made for distinctive texture on the top surfaces including the bottom/top surface of the channel (5). The height of each layer was 17.3 µm ± 2.4 µm (n=12) which is close to the 16 µm specified by the printer manufacturer and caused a layering effect in the sidewalls (4). The top cover for the closed channels had an organic, grown appearance at the edge (6) due to interaction between the support and build material. Channels fabricated by the DLP-SLA (c) were the closest in appearance to a device fabricated with traditional microfabrication methods, with straight sidewalls. The height of each layer was 37 ± 1 µm (n=3) with a roughness of 3 µm (7). The grid texture on the top surfaces, including inside the microchannel, appeared due to the pixels (56 x 56 µm) from the projector CCD and uneven illumination from each pixel. Back-exposure was a significant issue, as the closed channels would often block (8) when the resin trapped in the channel was exposed; careful optimization of the print parameters was required to avoid this. The correlation between the size of the designed and fabricated channels is shown in Fig.3. All of the printers demonstrated a linear correlation, however the accuracy and precision varied. Using FDM, it was impossible to fabricate laminar flow chips with channels smaller than 500 µm which correlates with finding by Lee et al. 22, and channels were consistently smaller (107 ± 36 µm) than designed; this is due to spreading of the polymer as it is extruded. It was observed that the diameter of the nozzle had significant impact on the channel width, however, it is possible to fabricate channels < 500 µm by reducing the distance between deposition passes as demonstrated by Anciaux et al.23 Fabrication by the PolyJet was found to produce slightly smaller channels (-40 ± 36 µm), with a similar deviation to the FDM. Closed and open channels as small as 205 ± 13 µm (designed 250 µm) were printed and successfully cleared of support material. The PolyJet was the only printer capable of reproducing channels of this size, however it struggled to fabricate positive features smaller than 200 µm which is typical of the technology.21

Figure 2. SEM micrographs of open and closed microfluidic channels. a) FDM structure where (1) Collapsed 250 µm and (2) 350 µm channels. (3) Deformed surface due to nozzle. b) PolyJet structure with (4) Layer thickness 16 µm (5) Surface texture due to scraper, (6) Deformed layers due to support interaction. c) DLP-SLA structure with (7) Layer thickness 32 ± 2 µm (8) Under-exposure causing partial collapse of channel, (9) Texture due to projector pixels 56 µm. Scale bars are 500 µm.


Page 5 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 3. Graph showing the relationship between CAD design and fabricated channels for the three printing technologies. PolyJet (diamond, black, large dashed line) shows a tendency to undershoot (40 ± 36 µm), FDM (square, red, small dashed line) also undershot by over one hundred microns (109 ± 37 µm), however DLP-SLA (triangle, green, solid line) demonstrated a tendency to overshoot (26 ± 20 µm); DLP-SLA demonstrated the closest to a direct replication of the original CAD design. All measurements were taken from SEM images using ImageJ. N=3.

The DLP-SLA printer was the only printer to produce channels slightly larger than the design (26 ± 20 µm) which correlates to half a pixel (56 x 56 µm), and was the highest in accuracy and precision. In addition the DLP-SLA was able to fabricate positive structures as small as 94 ± 7, typical of DLPSLA printers9,12. The superior accuracy and precision of the DLP-SLA makes this the better choice for microfabrication, however it cannot match the performance of traditional photolithography and soft-lithography processes.

4. The FDM (Ra=10.97 µm) top-surface shown in Fig. 4(a) contains a wide range in height across the surface, varying as much as 93 ± 5 µm. Valley like formations formed through the individual depositions of ABS were observed, with a peak-topeak distance of 603 ± 55 µm, which corresponds to the width of the nozzle diameter. The Polyjet produced a surface with smaller features (Ra=0.99 µm) but the texture of the roller used to level each deposited layer can been in Fig. 4(b). This leads to the formation of 5.8 ± 1 µm deep troughs and 1 µm (0.97 ± 1 µm) high fish bone like textures spaced every 147 ± 15 µm. Originating with the roller, these textures are typical of PolyJet or MultiJet printing21,22. The surface with the smallest features on the surface was obtained with DLP-SLA (Ra=0.35 µm, Fig. 4(c)). As already mentioned the grid pixel texture (57.5 ± 0.01 µm) was formed due to CCD pixel density, and resulted in a 0.73 ± 0.14 µm variation in height. The DLPSLA was found to produce the smoothest surface of the compared 3D printers, while FDM made the roughest. Production and infrastructure. Post-production for each of the three printers was different, with the times for each compared in Table 1. Post-production of FDM devices was simple, only removal of stray fibers from connectors was required immediately after printing, however as the layers left gaps between the port and the tubing, sealant/glue was needed to prevent leaking which took 1 h to fully cure. For the PolyJet, post-production was a significant issue as removal of the support material would take up to 3 days to remove from the channels. This is due to the requirement to solubilize the support in water, which was slow for long and narrow channels. This is the main restriction limiting the fabrication of very small channels < 150 µm channels and new support materials are needed if this is to be achieved.

Surface roughness. To obtain a more quantitative understanding of surface roughness, each device was analyzed with an optical profiler, with the surface profiler renders shown in Fig.

Figure 4. Surface roughness renders of FDM (a), PolyJet (b), and DLP-SLA (c). A) Major characteristics of the FDM are shown, with large valleys (peak- to-peak 602 ± 55 µm, ) formed due to the individual bead depositions of thermal polymer, and defined by the nozzle diameter (300 µm). B) Texture of the roller used to level each layer is shown, where 5.8 ± 1 µm deep troughs are formed, and repeated 1 µm high fish bone like textures, 0.97 ± 1 µm. C) Pixel texture (57.5 ± 0.01 µm) formed due to CCD pixel density, repeated throughout printed surfaces which was 0.73 ± 0.14 µm in height. Measurements taken using Vision 4.20. N=3.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 10

Figure 5. Microscope images of laminar flow within 500 x 500 into 750 x 500 µm channels, visualised with yellow and blue food dye at 25 µL/min for FDM, Polyjet, and DLP-SLA respectively. Below, are plots of distance vs. mixing ratio demonstrating diffusion through the laminar flow channel at 100, 50, and 25 µL/min respectively. A) The FDM (diamond, red, large dashed line) shows increased mixing due to the 300 µm wide valleys formed by the deposition of polymer beads pressing together; flow rate has little effect on the mixing rate with consistent mixing saturation at 15 mm. B) The Polyjet (square, blue, small dashed line) displayed a diffusion cone more like what would be expected in device of this kind fabricated with traditional methods. However, pinching can be observed at the origin, with increased mixing compared to the DLP-SLA, and FDM before stabilising. This was due to the layered 16 µm surfaces of the sideways causing increased mixing which was reduced at 100 µL/min and did not saturate, as did 50 µL/min. However the profile became similar to the FDM at 25 µL/min saturating at 15 mm. C) The DLP-SLA (triangle, green, solid line) showed stable low levels of mixing for all of the flow rates, with a limited range of error. As the flow rate decreased the level of mixing did also, however in a linear fashion with no saturation. N=3. Scale bar is 500 µm.

For DLP-SLA the post-production was minimal, only requiring bonding the device to a glass slide with resin and exposure to UV light. The simplicity allowed a device to be ready for use within 15 minutes. DLP-SLA possesses a strong advantage being able to fabricate channels with no support, but optimization of the resin for resolution, optical clarity, and back-exposure reduction needs considerable development. This has already been examined by Gong et al. were closed channels of 60 x 108 µm were achieved 11. Materials. The device materials in this work were selected for optical transparency over all other factors, as this is desirable for many microfluidic applications. Issues around biocompatibility have been investigated in other works, in general photopolymers being inherently more toxic due to residual initiator and monomers22,24-26. There are some biocompatible materials marketed for both SLA,20 and Polyjet27 printers, however these still show some toxicity without extensive treatment24. As FDM does not rely on photo-polymerization, it has a significant advantage over the DLP-SLA, and PolyJet technologies in terms of biocompatibility. However ABS is not always inherently biocompatible but can be made so by acetone based sealing and PEG grafting28. PLA would be a more suitable choice for biocompatibility29 although transparency of devices is an issue. Commercial translation of research devices into products requires translation of prototypes to injection molding, hot embossing, vacuum casting requires thermoplastics, making FDM attractive. Both FDM and Polyjet printers have the ability to print devices made from multiple materials, and

to blend these materials during the printing process. A wide variety of materials is available, with differences in physicochemical properties including, chemical resistance, thermal resistance, flexibility, electrical conductivity, magnetism, and porosity. Cost. The cost of the printers, both capital, consumable and time costs varies significantly as Table 1. At the time of writing, the approximate cost of the FDM printer used in this work, the Felix 3.0 is $1300 USD, the Eden 260VS is $85 000 USD, and the MiiCraft+ is $3000 USD. A sizable cost difference also exists in the cost of chips considering the cost of material, not upkeep of the printers, with the price per device $0.1 USD, $4 USD, and $2 USD for FDM, PolyJet, and DLPSLA, respectively. It is worth noting that these are still considerably cheaper than the $215 estimated for a new PDMS design15. The time to print a device was 6, 30, and 12 min for FDM, PolyJet, and DLP-SLA, respectively. It should be noted that the FDM, and DLP-SLA could only print one chip at a time due to technical issues. The FDM build platform was not uniformly level; this meant that the calibration of the zero position of the nozzle in the Z-axis, had to be established for different XY positions. For the DLP-SLA the limited XY (43 x 27 mm) build space, meant that the chip design used the entire XY build area; the alternative build orientation of placing the chip length-ways in relation to the Z-axis would have increased build time significantly. The maximal number of chips printed per hour was 10 for the FDM, 4 for the DLPSLA. In a single, two hour run, the PolyJet was able to fabri-


Page 7 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


cate 66 chips, providing the possibility to produce several hundred devices in a day, making it potentially attractive for small volume manufacturing. One should keep in mind, however, that the extensive post processing required for PolyJet printing currently limits the attractiveness of this option for same day prototyping. To summarize, FDM provides chips at a low cost at the expense of resolution, while the Polyjet provides the inverse. DLP-SLA offers high resolution at the expense of throughput, with a cost significantly below than of the Polyjet. Laminar flow chip design. To examine the fluidic properties of the printed devices, a simple Y junction design was selected for laminar flow studies and fabricated with each of the 3D printers. This design was selected in order to determine the optimized settings for each of the 3 printers as it can be used for mixing, reaction, and parallel two phase flow. The width and shape of the diffusion area formed between the two flows was examined as a way to determine differences in microfluidic behavior for the three print methods. While each of the designs had the same internal channel dimensions, careful consideration of the different methods of 3D printing was considered to optimize the device for each printer type. The footprint of the chip was chosen to be 40 x 20 mm as the max XY build space of the MiiCraft+ was 43 x 27 mm. The minimum channel size included in the comparison was 500 µm, limited by the smallest dimension that could be printed on the FDM printer. PolyJet design. We observed that the VeroClear polymer was brittle and prone to cracking when less than 1 mm in thickness, hence devices with a thickness over 1 mm were printed with the PolyJet to retain strength and ease of handling. While brittle at the smaller thickness, once over 1 mm in thickness, the chips were difficult to snap or bend. Vertical connectors were designed with straight sidewalls, 100 µm less in diameter less than the connecting PEEK tubing (1.6 mm) to provide a tight fit. The rough sidewalls due to the 16 µm layer deposition allowed fluidic sealing at the flow rates examined. As the negative minimum feature size for the Polyjet was found to be 205 ± 13 µm, rulers were included in the design to aid with mixing measurements. FDM design. With the FDM device, the channel was difficult to visualize once multiple layers of polymer were deposited under the channel to form the chip base. This was due to the cross-hatch nature of the base and entrapment of air between the filament, due to incomplete fusing of the extruded polymer. The optical clarity was improved by reducing the base layer to a single 100 µm layer as the ABS was strong enough to be removed from the build platform without deforming or causing leaks. In addition, due to the strength of the ABS, it was possible to reduce the material in the device, saving on material and print time. When forming the 500 µm channels, the printer would direct the nozzle across the channel providing a risk of a small bead of ABS blocking the channel. To overcome this, the walls of the chip were designed to be exactly 3 layers of ABS wide (900 µm), and the loop pass number in the software was increased to 3. Therefore the channels were formed by continuous movements of the nozzle head, with no crossover. The inlets of the devices printed by FDM were designed to address two problems. First, by reducing the height of the device the print time per device was reduced from 30 to 10 min. Second, the inlets into which tubing was

connected were horizontally orientated, reducing the effective gap around the inserted tubing and improving sealing to the tubing. This also avoided the circular structures required for vertical connection ports, which normally are fabricated in a spiral, leaving a gap equal to the nozzle head diameter. DLP-SLA design. Printing using the DLP-SLA was most difficult to optimize of the three printers. Variables in fabrication conditions (room temperature, exposure times, Teflon film position, slicing software, resin age) all impacted on the outcome. The final design was reached for four reasons. First, by making the bottom layer of the chip relatively thick (500 µm), the structural stability of the device was sufficient to provide the robustness required for easily handling. A minimum of 4 layers (or 200 µm) was required to seal the channels and avoid back-exposure into the channel. Second, the fluidic connector went through numerous design iterations minimizing back-exposure in both vertical and horizontal configurations. The final design involved tapered connectors from 5 mm to 2.5 mm, and removing the section directly above the channel. Third, as the sidewall roughness was < 1 µm, sealing of the tubing without sealant was an issue. To address this, a collar with a reduced diameter of 1.4 mm was introduced to grip the tubing and seal the connector. Fourth, the forces between the cured layer and Teflon film caused devices to stick to the build space; this effect increases with surface area. Therefore by reducing the surface area by removing sections of the bulk next to the laminar flow channels, we prevented sticking while still providing structural integrity of the device. Microfluidics – laminar flow and mixing. The suitability of the chips fabricated by the three 3D printers for microfluidics was studied. Laminar flow profiles at a flow rate of 25 µL/min are shown in Figure. 5 of FDM (a), PolyJet (b), and DLP-SLA (c), respectively. The extent of mixing for flow rates of respectively 100, 50, and 25 µL/min are shown below the flow photos for each printer. FDM microfluidics. In the FDM chips, the two dyes started flowing parallel, but mixed rapidly with complete mixing achieved within 15 mm from the intersection for all flow rates. The large peaks and valleys formed through the fabrication process (Fig. 4) cause additional flow interfaces with folds and stretches as described by Simonnet and Groisman30. This mixing occurs when the flow from one of the inlets was parallel to the printed grain in the chip. Considering the popularity of surface patterning for the creation of passive micromixers, FDM may be a simple and effective way to make micromixers. Alternatively, their use should be restricted to applications where mixing has no impact on the outcome. PolyJet microfluidics. In the PolyJet chips, the fluids reached complete mixing at 25 mm for 50 µL/min, and 15 mm for 25 µL/min. Unexpectedly mixing was observed at the origin which we believe is caused by the 16 µm channel structures present in the sidewalls running parallel to the flow profile (Fig. 5). At high flow rates (100 µL/min), this effect was reduced and mixing was 83 ± 6% at 25 mm. At 25 µL/min, however, more mixing was observed at 5 mm than when using FDM printed devices (82 ± 6% vs 71 ± 12% for PolyJet and FDM, respectively). Microfluidic devices where complex geometries are required without the need for precision in channel reproduction such as droplet or cell culture systems are most suited to PolyJet type printers.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

DLP-SLA microfluidics. The DLP-SLA chips showed the least mixing, and complete mixing was not achieved at any of the flow rates shown (complete mixing only occurred at 1 µL/min). The highest extent of mixing was 32 ± 7% at 25 mm and 25 µL/min. These results are in agreement with the low surface roughness (Fig. 4), high design reproducibility (Fig. 3) and provides flow profiles similar to those observed in more conventionally fabricated devices 31. Microfluidic devices that require laminar flow and minimization of mixing, for example, diffusion based H-filters, are most suited to be made with these types of printers. Conclusions Comparing the performance of the three most popular 3D printing technologies used to make microfluidic devices, we found that they all have significant advantages. While microchannel resolution in general is still a major issue for all 3D printers, each of the different types of printers has unique properties when it comes to making microfluidic devices. FDM printers are the cheapest, to both purchase and to use, and have the largest selection in material choice – some of which are biocompatible (PLA, COC), electrically conductive (composites with carbon, cooper, graphene), and specific chemical resistance (COC for example which is resistant to hydrolysis, acids, alkalis, as well as polar solvents such as methanol). Its reliance on thermoplastics inherits compatibility with traditional polymer manufacturing potentially, helping the translation to commercial products. Multiple material printing is another strong advantage for FDM, which is also shared by the PolyJet. While low resolution (321 ± 5 µm) particularly in the XY plane means that < 500 µm is an impossibility at this stage, the roughness (10.97 µm) does make the FDM well suited to fabricating low cost micromixers. The Polyjet printer was able to produce the smallest closed channels (205 ± 13), and was capable of producing the largest number of chips per hour (33 chips), making the Polyjet a potentially attractive option for low volume manufacturing in research environments. However the difficulty and time needed in removing support complicates post processing which is significant disadvantage. In addition with the high-initial purchase price of the printer and high consumable cost, it has the highest cost per device of the three printers compared. We see the Polyjet being useful for fabricating complex microfluidic systems for droplet generation, and cell culture platforms. The resolution of the DLP-SLA allows for the smallest open channels (154 ± 10), with well-defined channels, and fast postprocessing. The major advantage for the DLP-SLA devices was the laminar flow performance - fluidic behavior is similar to what is achieved in current microfluidic devices - however the small XY build space (43 x 27 mm) limits the throughput, making it mainly suitable for single-device prototyping. Like the PolyJet, DLP-SLA relies on the use of commercial photopolymers, inherently compromising biocompatibility. The primary disadvantage of the DLP-SLA is that due to not using support, back exposure of channels limits closed channel fabrication to 500 µm with our printer. These results indicate the DLP-SLA is well suited to microfluidics were precise control of features and fluidics is critical. The selection of the ‘best printer for the job’ requires careful consideration of the requirements of the device, and application; compromises will have to be made on some aspects as

Page 8 of 10

there is no technology covering all microfluidic applications. We believe that to ensure that tomorrows printers will better cover demands specific to the microfluidic community, communication of these demands through partnerships and conversation with 3D printing engineers are required.

ASSOCIATED CONTENT Supporting Information The Supporting Information includes the .stl files of the 500 µm channel devices (FDM, PolyJet, DLP-SLA) designed, fabricated, and tested in this work. The Supporting Information is available free of charge on the ACS Publications website. Supplementary Information (SI) (.RAR) – including: FDM laminar flow device (.stl) PolyJet laminar flow device (.stl) DLP-SLA laminar flow device (.stl)

AUTHOR INFORMATION Corresponding Author * Prof. Michael C. Breadmore, ARC Centre of Excellence for Electromaterials Science, School of Physical Sciences, University of Tasmania, Sandy Bay, Hobart 7001, Tasmania, Australia.

Author Contributions N.P.M. designed, fabricated, performed experiments, collected data, and prepared the manuscript; J. M. C. and P.S. assisted with the experiments, and paper editing; R. G., M.C.B. and B.P. conceived and supervised the research and edited the paper. All authors have given approval to the final version of the manuscript.

ACKNOWLEDGMENT Dr. Sandrin Feig is thanked for his assistance in SEM imaging. This study is supported by ARC Centre of Excellence for Electromaterials Science (ACES) (Grant CE140100012). The authors thank the Central Science Laboratory (CSL), MCB would like to thank the ARC for Future Fellowship (FT130100101).

REFERENCES (1) Sahore, V.; Kumar, S.; Rogers, C. I.; Jensen, J. K.; Sonker, M.; Woolley, A. T. Anal Bioanal Chem 2016, 408, 599-607. (2) Hull, C. W.; Google Patents, 1986. (3) Mukhopadhyay, R. Anal. Chem. 2007, 79, 3248-3253. (4) Pham, D. T.; Gault, R. S. Int. J. Mach. Tool. Manu. 1998, 38, 1257-1287. (5) Vaezi, M.; Seitz, H.; Yang, S. Int. J. Adv. Manuf. Tech. 2013, 67, 1721-1754. (6) Waheed, S.; Cabot, J. M.; Macdonald, N. P.; Lewis, T.; Guijt, R. M.; Paull, B.; Breadmore, M. C. Lab Chip 2016, 16, 19932013. (7) McDonald, J. C.; Chabinyc, M. L.; Metallo, S. J.; Anderson, J. R.; Stroock, A. D.; Whitesides, G. M. Anal. Chem. 2002, 74, 1537-1545.


Page 9 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(8) Symes, M. D.; Kitson, P. J.; Yan, J.; Richmond, C. J.; Cooper, G. J. T.; Bowman, R. W.; Vilbrandt, T.; Cronin, L. Nat. Chem. 2012, 4, 349-354. (9) Shallan, A. I.; Smejkal, P.; Corban, M.; Guijt, R. M.; Breadmore, M. C. Anal. Chem. 2014, 86, 3124-3130. (10) Rogers, C. I.; Qaderi, K.; Woolley, A. T.; Nordin, G. P. Biomicrofluidics 2015, 9, 9. (11) Gong, H.; Beauchamp, M.; Perry, S.; Woolley, A. T.; Nordin, G. P. RSC Adv. 2015, 5, 106621-106632. (12) Gong, H.; Woolley, A. T.; Nordin, G. P. Lab Chip 2016, 16, 2450-2458. (13) Wehner, M.; Truby, R. L.; Fitzgerald, D. J.; Mosadegh, B.; Whitesides, G. M.; Lewis, J. A.; Wood, R. J. Nature 2016, 536, 451-455. (14) Sandron, S.; Heery, B.; Gupta, V.; Collins, D. A.; Nesterenko, E. P.; Nesterenko, P. N.; Talebi, M.; Beirne, S.; Thompson, F.; Wallace, G. G.; Brabazon, D.; Regan, F.; Paull, B. Analyst 2014, 139, 6343-6347. (15) Au, A. K.; Huynh, W.; Horowitz, L. F.; Folch, A. Angew. Chem., Int. Ed. 2016, 55, 3862-3881. (16) Bhattacharjee, N.; Urrios, A.; Kang, S.; Folch, A. Lab Chip 2016, 16, 1720-1742. (17) Chen, C.; Mehl, B. T.; Munshi, A. S.; Townsend, A. D.; Spence, D. M.; Martin, R. S. Anal. Methods 2016, 8, 6005-6012. (18) Ho, C. M. B.; Ng, S. H.; Li, K. H. H.; Yoon, Y. J. Lab Chip 2015, 15, 3627-3637. (19) Reza, A.; Stephanie, K.; Alexander, H.; Bekir, Y.; Fariba, G.; Sara, K.; Michael, M.; Ali, K.; Savas, T. Biofabrication 2016, 8, 022001.

(20) Au, A. K.; Lee, W.; Folch, A. Lab Chip 2014, 14, 12941301. (21) Rafał, W.; Krzysztof, A. J. Micromech. Microeng. 2015, 25, 085013. (22) Lee, J. M.; Zhang, M.; Yeong, W. Y. Microfluid. Nanofluid. 2016, 20, 5. (23) Anciaux, S. K.; Geiger, M.; Bowser, M. T. Anal. Chem. 2016, 88, 7675-7682. (24) Macdonald, N. P.; Zhu, F.; Hall, C. J.; Reboud, J.; Crosier, P. S.; Patton, E. E.; Wlodkowic, D.; Cooper, J. M. Lab Chip 2016, 16, 291-297. (25) Oskui, S. M.; Diamante, G.; Liao, C.; Shi, W.; Gan, J.; Schlenk, D.; Grover, W. H. Environ. Sci. Technol. Lett. 2016, 3, 1-6. (26) Urrios, A.; Parra-Cabrera, C.; Bhattacharjee, N.; GonzalezSuarez, A. M.; Rigat-Brugarolas, L. G.; Nallapatti, U.; Samitier, J.; DeForest, C. A.; Posas, F.; Garcia-Cordero, J. L.; Folch, A. Lab Chip 2016, 16, 2287-2294. (27) Sayyar, S.; Murray, E.; Thompson, B. C.; Chung, J.; Officer, D. L.; Gambhir, S.; Spinks, G. M.; Wallace, G. G. J. Mater. Chem. 2015, 3, 481-490. (28) McCullough, E. J.; Yadavalli, V. K. J. Mater. Process. Technol. 2013, 213, 947-954. (29) Athanasiou, K. A.; Niederauer, G. G.; Agrawal, C. M. Biomaterials 1996, 17, 93-102. (30) Simonnet, C.; Groisman, A. Phys. Rev. Lett. 2005, 94, 134501. (31) Hessel, V.; Löwe, H.; Schönfeld, F. Chem. Eng. Sci. 2005, 60, 2479-2501.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 10

For TOC only

10 ACS Paragon Plus Environment

Comparing Microfluidic Performance of Three-Dimensional (3D

Recommend Documents